Cellular and Colloidal Separation Using Optical Forces

CHAPTER 17

Cellular and Colloidal Separation Using Optical Forces Kishan Dholakia,* Michael P. MacDonald,* ˇ izˇma´r† Pavel Zema´nek,† and Toma´sˇ C *SUPA, School of Physics and Astronomy University of St. Andrews, Fife, KY16 9SS Scotland †

Institute of Scientific Instruments ASCR, v.v.i. Academy of Sciences of the Czech Republic 61264 Brno, Czech Republic

I. Introduction II. A Brief Review of Fluorescence-Activated Cell Sorting (FACS) and Magnetically Activated Cell Sorting (MACS) III. Optical Forces for Cell and Colloidal Sorting: Theoretical Considerations IV. Overview of Experimental Optical Force-Based Sorting A. Active Sorting with Optical Forces B. Passive Sorting with Optical Forces V. Flow-Free Optical Methods VI. Optical Methods with Flow A. Optical Chromatography B. Separation Using Microfluidic Flow Over Periodic Optical Energy Landscapes VII. Dielectrophoresis for Microfluidic Sorting VIII. Conclusions References

The separation or sorting of cellular and colloidal particles is currently a central topics of research. In this chapter, we give an overview of the range of optical methods for cell sorting. We begin with an overview of fluorescence and magnetically activated cell sorting. We progress to describing methods at the microfluidic scale level particularly those exploiting optical forces. We distinguish between METHODS IN CELL BIOLOGY, VOL. 82 Copyright 2007, Elsevier Inc. All rights reserved.

467

0091-679X/07 $35.00 DOI: 10.1016/S0091-679X(06)82017-0

Kishan Dholakia et al.

468

what we term passive and active schemes for sorting. Optical forces pertinent to the sorting schemes are described, notably the gradient force and the optical radiation pressure (or scattering force). We discuss some of the most recent advances. This includes techniques without fluid flow where we have either stationary or moving light patterns to initiate separation. Further methods have shown how using an externally driven flow either counter-propagating against a light field (optical chromatography) or over a periodic light pattern (an optical potential energy landscape) may result in the selection of particles and cells based on physical attributes such as size and refractive index. We contrast these schemes with the field of dielectrophoresis where electric field gradients may separate cells and also briefly mention the upcoming area of light-induced dielectrophoresis which marries the reconfigurability of optical fields with the power of dielectrophoresis.

I. Introduction The use of the modern optical microscope in unison with laser technology has fueled a revolutionary advance in cellular and molecular biology. A number of microscopic methods have allowed scientists to view how proteins and lipids behave and how their interactions govern the intricate mechanics, maintenance, and function of the intracellular world. Real-time observation and tracking of cellular processes with appropriate fluorescent tagging has yielded a wealth of bioscience (Prasad, 2003). In parallel with the outstanding advances in microscopy, there have been other complementary advances in the use of laser light at the cellular scale. Most importantly, there has been the exploitation of the forces of light to trap, cut, move, and, more recently, sort biological and colloidal materials. As a key example, optical tweezers (Ashkin and Dziedzic, 1987) allow micron-sized particles and cells to be trapped, moved, and generally manipulated without any physical contact and is discussed in other chapters of this volume. This area of research, more broadly termed optical micromanipulation, is undergoing a growth of activity at the current time with a particular emphasis in the biological and colloidal sciences. In biology, this methodology has revolutionized our understanding of molecular motors. Rather than trapping just one or two objects at a time, there is interest in creating an array of trap sites. Such multiplexed optical trapping may be implemented by the use of holographic, interferometric, and acousto-optic devices (Grier, 2003; Molloy et al., 2003; Neuman and Block, 2004). Such multiple optical traps may create what is termed as an optical potential energy landscape. This chapter deals with the concept of sorting of cells and colloidal samples. Particle motion on optical landscapes is central to the newly emerging forms of separation. At the microscopic scale and nanoscale, many disciplines seek methods for accurately and eYciently separating colloidal, cellular, and other biological particles. Such selection plays a pivotal role in enabling studies in biology and medicine. As an example, at the nanoscale the technique of gel electrophoresis permits sorting of DNA by size. The isolation of specific cell subpopulations is central

17. Cellular and Colloidal Separation Using Optical Forces

469

to the advancement of cell-based therapies for cancer, autoimmune diseases, and genetic disorders. This includes the ability to select stem cells from a population. Stem cell populations are key to important areas in modern medicine—in regenerative medicine, sources of stem cells can be exploited to provide new disease-free tissue. Tumor stem cell populations are key for disease development and for successful therapy. Methods for separating and investigating stem cell populations are thus fundamental to providing a rational basis for improved disease understanding and new forms of therapy. Excellent macroscopic schemes exist to perform high-throughput, multiparameter cell separation, typically based on the conventional flow cytometer. High yield in such a system is typically realized with inputs in excess of 105 cells/s. However, the ability to miniaturize a cell sorting system and move toward a microfluidic basis for this technology, we may gain some advantages. Smaller number of cells may be separated and yet deliver a high yield. Reagent use is dramatically limited in such a small environment and the method may find favor with rare or precious cells, for example primary cells that may not lead to large cell populations. If the device is small and inexpensive, it may oVer a disposable, sterile platform for cell separation that would bring the technology of cell separation to a wider number of researchers in the biosciences. This chapter focuses on some of the emergent methods for cell and colloidal particle sorting at the microscale using minimal volumes of analyte and employing optical forces. Much of this work has been performed in only the last 5 years and it is a young and vibrant area of research. We cannot hope to cover the whole of this field in depth in this chapter, but believe we may give the reader a flavor of the work being performed, the shortcomings and challenges that lie ahead, and a representative list of references. We begin by reviewing established methods for cell sorting using macroscopic apparatus with large throughputs and discuss how they are being miniaturized. We focus on the use of optical forces to separate cells with or without markers attached to them, distinguishing what we term passive sorting from active sorting. We then review sorting techniques both with and without a microfluidic flow present. We conclude with a short comparison to the method of dielectrophoresis for cell sorting at the microscale.

II. A Brief Review of Fluorescence-Activated Cell Sorting (FACS) and Magnetically Activated Cell Sorting (MACS) The fluorescence activated cell sorter first appeared in the late 1960s and is based on a flow cytometer. This device records the properties of a single cell as it traverses a laser beam. In doing so, the cell scatters light and this is then recorded on suitably placed detectors. A FACS machine is a special type of flow cytometer: fluorescent markers (fluorescently labeled monoclonal antibodies) allow specific cells to be recognized and subsequently separated. Analysis and separation based on a wide range of parameters may be implemented. A review by the inventors of FACS in 2002

Kishan Dholakia et al.

470 Sample fluid

Sheath fluid

Hydrodynamic focusing Scattering detection

Light sources

Charging and droplet formation

Fluorescence detection

+ − − + − +

Sorted cells

Waste

Sorted cells

Fig. 1 Diagram of FACS machine. Cells have been fluorescently tagged (shown as black or white), with some cells remaining untagged. On the basis of the fluorescence signal detected, a charge is applied to individual droplets such that they can be deflected into separate chambers.

stated that FACS can record 12 diVerent fluorescent colors and 2 scattering parameters, which could all be used for sorting (Herzenberg et al., 2002). The technique can measure the cell size, volume, or viscosity. DNA or RNA content as well as the presence of surface antigens. FACS has a variety of uses and has been used in the diagnosis of leukemia, lymphoma, immunodeficiencies, and for compatibility in transplants. Figure 1 shows a schematic of a cell sorting device. Air pressure pushes cells out of a nozzle at high speed: a liquid jet (e.g., saline) is combined with the cells and acts as a sheath flow. An acoustic vibration is coupled to the nozzle’s tip, leaving a trail of cyclical imprints onto the liquid’s surface. Surface tension pulls at the waists

17. Cellular and Colloidal Separation Using Optical Forces

471

between the imprints, forcing the jet to separate into regularly spaced droplets. The cells, first carried by a cylinder of liquid, are thus distributed among a string of discrete droplets. After the cell leaves the nozzle region, it passes through the waist of one or more tightly focused laser beams which are at appropriate wavelengths to perform scattering or fluorescence excitation of the cell markers. The scattered and fluorescence light from these interactions is collected and analyzed. If a cell meets one or more criteria set by the user, an electrical charge may be applied to the droplets containing cells of interest as they separate from the main jet. In this manner, droplets with diVerent cell types are directed toward separate collection vials by a static electrical field (Fig. 1). The speed of modern electronics and detection systems means one may easily analyze cells at speeds greater than 105 events per second. Cell separation rates of the order of 10,000 per second are standard. We refer the reader to the excellent texts on fluorescence-activated cell sorting for further reading (Givan 2001). Fluorescence-activated cell sorting (FACS) is very powerful and may discriminate cells within a large parameter space; however, in several instances, one may just wish to separate two sets of cells quickly, and at less expense, than a traditional FACS machine. Here magnetically activated cell sorting (MACS) is worth considering. For immunomagnetic separation, cells are incubated with paramagnetic microbeads that are coated with appropriate antibodies. Subsequently, the cells of interest or the unwanted cells may be sorted by use of one or more magnets in a suitable array. The magnetic method is, to some extent, limited by the constraints one places on the parameters for separation and the obvious need for suitable antigens on the cell surface for the paramagnetic beads to bind to. The method is useful for rapid bulk separations or as a precursor to more elaborate sorting schemes. In the next sections, we turn our attention to recent work using optical forces to instigate cell and colloidal separations at the microscopic scale. We begin by exploring the optical forces which are relevant to sorting and the role of a microfluidic flow therein.

III. Optical Forces for Cell and Colloidal Sorting: Theoretical Considerations We distinguish four key regimes for the optically mediated sorting process at the microfluidic scale, which are represented in the phase diagram shown in Fig. 2. Later in this chapter, we will look at experiments with emphasis particularly on phases 1–3. Region (1): Static fluid, static pattern: light-induced flow and separation of microobjects The most basic form of separation exploits the diVerences in the aYnity of colloidal and biological particles to a stationary optical potential energy landscape with no fluid flow present. Any affinity diVerences would manifest themselves as

Kishan Dholakia et al.

Pattern movement

472 3 Static fluid Dynamic pattern

4 Dynamic fluid Dynamic pattern

1 Static fluid Static pattern

2 Dynamic fluid Static pattern

Fluid flow

Fig. 2 The dynamic phases available for optical separation. As the analyte can be either static or dynamic, in the presence of a static or a dynamic optical pattern, we have four diVerent phases within which to work.

particle sorting and separation in the absence of flow. Motion of objects can be investigated and this is achieved solely by optical forces. How fast the particle moves over the landscape depends on the energy depth of the potential landscape well (trap). The time a particle needs to overcome a potential barrier via Brownian motion to the particle size via Brownian motion. Directed particle motion was obtained by tilting the periodic optical landscape. Overall, this means that particles of diVerent size, shape, refractive index, and composition move diVerently across the optical landscape. For example, we shall see that red and white blood cells may be separated by placing them in a tailored circularly symmetric light beam pattern or spherical objects of desired size placed into a three-beam interference field will be transported in opposite directions (Zema´nek et al., 2004b). Region (2): Dynamic fluid, static pattern: microfluidic optical sorting Flowing particles through an optical potential energy landscape [two-dimensional (2D) or 3D interference pattern (a lattice), holographically produced array, timeshared array] will facilitate the separation of these objects based on their physical characteristics such as size, shape, and refractive index. Flow through an optical lattice can readily lead to sorting as a function of size and/or refractive index-related deflection (MacDonald et al., 2003; Pelton et al., 2004). Region (3): Static fluid, dynamic (moving) pattern By creating a time-varying optical landscape (e.g., scanning the beam to create, for example linear interference fringes moving perpendicular to the fringe axis), we can transport particles as if on a mechanical conveyor belt and use the fact that diVerent particle species respond diVerently. From a physical perspective, it is another way by which we can introduce a tilted or biased optical landscape. Naturally, this sensitivity of a given particle size on the landscape varies as a function of particle size and

17. Cellular and Colloidal Separation Using Optical Forces

473

relative refractive index between the object and the media. The traveling landscape provides a convenient means of pumping and actuation, but the distance the particles are carried will depend on many factors, including the shape and extent of the envelope of the optical pattern itself: as the intensity in the pattern falls oV more strongly, interacting particle species will be carried further than those that respond weakly to the optical forces. Since this dynamic phase of operation does not have a fluid flow to remove sorted particles or cells, particles will accumulate at the position for which the optical conveyor belt is no longer strong enough to transport them, leading to a fractionated column of diVerent particle species which may then be extracted. Region (4): Dynamic fluid, dynamic pattern, microfluidic sorting with enhanced particle separation The final regime utilizes a motional light pattern within a microfluidic flow. To date, this is a little explored regime even though it oVers more degrees of freedom with respect to tuning the sorting process. A particular advantage of this combination is that deflected particles can be moved out of the polydisperse flow more quickly, leading to fewer particle–particle interactions that might otherwise lead to undesirable behavior of the particles (e.g., clustering, jamming, incomplete deflection, and/or separation). This will allow higher eYciencies to be achieved in the sorting process, something that may be key to future implementations of all optical sorting. As a precursor to looking at the techniques being developed in all four regimes described above, we will now discuss the behavior of particles within an optical trap. Optical forces generate a mechanical eVect on atoms, molecules, and particles right up to the size of microscopic colloidal particles and single cells. An optical trap may readily generate piconewton forces in a noninvasive manner on cellular and colloidal particles. A particle in a trap behaves as a highly overdamped harmonic oscillator with a stiVness in the range of 0.05 pN/nm (Molloy et al., 2003; Neuman and Block, 2004). The object may escape from a trap due to thermal activation and the dynamics of motion may change in the presence of flow. This issue of thermal activation was studied over 60 years ago, Kramers (1940) elucidated the dynamics of particles in a double-well potential which can be approximated by closely spaced optical traps. It was shown that the mean time (Kramers time) tK to get the object over an energy barrier of height DU can be described by an exponential law of the form tK ¼ RexpðDU=kB TÞ, where T is the temperature, kB the Boltzmann constant, and R is a quantity depending on the potential curvature at the maximum and minimum. Microscopic colloidal particles or cells in an optical potential represent a powerful means by which to study such activation. Two components of optical force can be distinguished at the nanoscale. The gradient force is dependent upon which may scale linearly with the particle volume and is dependant upon the particle polarisability. This pushes the object to the higher (lower) intensity place if its refractive index is higher (lower) than the surrounding

474

Kishan Dholakia et al.

medium. The scattering force is proportional to the square of the particle volume and pushes the object in the direction of propagation of the incident light. Therefore, the gradient force spatially localizes the object while the scattering force pushes it away along the beam propagation axis. The object is trapped at equilibrium, (a position of zero total force) and any small object deviation from this position results in restoring force proportional to displacement. Since the bigger object is strongly pushed out of the trapping region of the gradient force, trapping can be reached only in tightly spatially localized field intensity maxima. In contrast, the optical force has a weak dependency on the object size, if the trapped object has dimensions one order larger than the wavelength. Straightforward single beam trapping is a possible method of optical sorting, if we select cells from a microfluidic environment and divert them into the channel of interest. The use of extended optical potential energy landscapes (periodic light patterns) is also of interest. Such periodic patterns may trap cells and colloid in the bright parts of the light field, if they are smaller than the pattern period and of higher refractive index than the surrounding medium. However, if the objects become larger than the pattern period or pitch they could settle with their center in the dark parts of the field. Intermediate-sized objects may not even sense the presence of the periodic field pattern and stay largely unaVected by the optical landscape. This, in turn, implies that sorting of objects is based on their size and polarisability. We concentrate here on particles in extended optical potential energy landscapes. In one dimension, particle behavior in an optical-standing wave was theoretically studied by Zemanek and colleagues for nanoparticles (Zema´nek et al., 1998) and microparticles showing a size-dependent eVect (Lekner, 2005; Siler et al., 2006; Zema´nek, 2002, 2003). They showed that the sensitivity of a spherical object of radius a to the periodic light pattern of period L can be described analytically for a weakly polarized particle (with its refractive index close to that of the surrounding medium) using the force acting on it perpendicular to the fringes: 

    2pa 2pa 2pa 2pðx  x0 Þ F ðxÞ
F0 sin  cos sin L L L L

ð1Þ

where x0 is the position of the closest peak to the beginning of axis x. The first square bracket (size term—see Fig. 3) has oscillatory behavior passing through zero at a/L
0, 0.715, 1.230, 1.735,. . . and having maxima and minima separated by L/2 starting at amax ¼ L/2. Therefore, if the size term is positive (e.g., particle radius is smaller than 0.715 L), the sphere center is localized at x ¼ x0 þ ML, where M ¼ 0, 1, 2,. . . This position corresponds to the intensity maximum of the fringe. In contrast, if the size term is negative, then the equilibrium position is shifted by L/2 and the sphere center is positioned at the intensity minimum of the fringe. If the size term is close to zero, no matter where the sphere is localized in the periodic pattern, the force equals to zero and the sphere does not feel the periodic landscape. The above-mentioned particle sensitivity to the periodic pattern (that may on

475

17. Cellular and Colloidal Separation Using Optical Forces

10

Size term

5 0 −5 −10 −15

0

0.5

1 a/Λ

1.5

2

Fig. 3 Plot showing how the size term sinð2pa=LÞ  cosð2pa=LÞ2pa=L in Eq. (1) depends on the size of the sphere. This dependence is exactly valid only if the refractive index of the object is close to that of the surrounding medium. But the qualitative behavior has more general validity.

occasion be more complex than described here) underpins all four types of optical sorting in periodic light patterns (Fig. 2). Sorting of nanoparticles according to their optical properties was studied theoretically in a 1D pattern obtained using a three-beam configuration (Zemanek, 2004a). The same field configuration was used later to analyze static sorting of microspheres according to their size or optical properties (Zemanek, 2004b). The particle separation was based on the opposite light-induced particle flow if the particle center settles in a fringe maximum or minimum. The size eVect also plays an important role if the sorting is based on thermally activated jumps over a barrier between neighboring equilibrium positions. The lower the barriers, the more frequently a particle will jump out of a given trapping region. This facilitates more rapid motion along the tilted periodic landscape (Reimann, 2002; Tatarkova et al., 2003). In separate studies, nanoparticle sorting using a plasmon resonance excitation has been studied theoretically (Zelenina et al., 2006). Turning now to sorting in the presence of flow, Pelton et al. (2004) presented a theoretical model that took into account the Brownian motion of objects smaller than the trapping wavelength in 2D periodic potentials. Here one needs to examine the eVect of the optical forces competing with the Stokes forces (viscous drag) as the external driving term. For such small objects, they approximated the optical forces acting on them and expressed the analytical solution for deflection angles in separable 2D landscapes. The particle motion may be described by a Langevin equation and particles may be locked-into certain directions that are not correlated with the direction (orientation) taken by the fluid flow. They explored the sensitivity of various landscapes to parameters such as size and theoretically showed that, in general, extended periodic landscapes would yield excellent sorting resolution and exponential size selectivity. Alternative theoretical studies including general 2D potentials and the influence of thermal noise have been presented by Lacasta and colleagues (Gleeson et al., 2006; Lacasta et al., 2005). They used the Langevin equation

Kishan Dholakia et al.

476

A

B 6

C

P(T)d 2T

4

5 mm

D

2

0

Fig. 4 Kinetic lock-in of colloid flowing through a static optical potential landscape. Depending on the orientation of the landscape to the flow diVerent behavior is seen. Relative probability that a sphere will pass through a point in the field of view, when the direction of the trap lattice is oriented at (A–D). In all figures, the external flow is from left to right. As the behavior of particles with diVerent size or refractive index will diverge, this and similar phenomena were later used to obtain sorting of colloid and cells. Reprinted figure with permission from Korda, P. T., Taylor, M. B., and Grier, D. G. (2002). Phys. Rev. Lett. 89, 128–301, copyright (2002) by the American Physical Society.

to study particle behavior in these potentials and they especially looked for particle trajectories, if they were driven over the periodic structure under diVerent angles and various force amplitudes (Lacasta et al., 2005). This driving force may be exerted, for example, by fluid flow (viscous drag) as previously mentioned. They observed a terrace phenomenon in the dependence of the absolute velocity angle on the direction of the force. This phenomenon has been observed experimentally by Korda et al. (2002) (Fig. 4) and termed kinetically locked-in states. Gleeson et al. (2006) presented an analytical approach to this phenomena based on the overdamped Fokker–Planck equation. They derived a first-order approximation to the average velocity vector v in the form v¼Fþ

ð

1 ð2pÞ

4

k2 K ^ ðkÞ; dK
Q 2 Tk  kF

ð2Þ


is the dimensionless where F is the vector of the uniform driving force, T ^ kÞ is defined by Q ^ ðkÞ ¼ V ^ ðkÞV ^ ðkÞ, where V ^ is the Fourier temperature, and Qð transform of the potential V ðxÞ (periodic or even random) and k2 ¼ k  k. The

17. Cellular and Colloidal Separation Using Optical Forces

477

validity of this approximation is fairly good, especially for higher temperatures and forces.

IV. Overview of Experimental Optical Force-Based Sorting Techniques which use optical forces to sort cells can be placed into one of two groups. We call the first group of techniques active sorting, where an external decision based on probing a particle passing a detection region is part of the sorting mechanism. We have already seen examples of this in our brief discussions of FACS and MACS. The second group may be termed passive sorting, where the selection and separation of cells occur without the need for any external decisionmaking process, but rather are based on the diVering aYnity of a given object to the light field in the presence of an external drive (the fluid flow). Both of these groups allow for the attachment of markers which give enhanced selectivity, but it is predominantly the active group that relies on these markers while the passive techniques are usually promoted as needing no such markers or tags. It is also instructive to consider some of the fluid dynamics that will be pertinent to microfluidic cell sorting: at the microfluidic scale, fluid flow is what we term laminar: we are typically in the low Reynolds number regime. The Reynolds number is the ratio of inertial to viscous forces: thus, we are ignoring inertia to a large extent and are reliant solely on viscous forces: turbulence is not present and thus for tasks such as mixing, deflection, and sorting we have to rely, in the absence of optical or other forces, solely on diVusion. Thus, we may have two fluid streams running parallel to one another with little or no mixing taking place. In terms of sorting, as this means for such a geometry, our optical forces may deflect particles of choice readily from one flow to another (Squires and Quake, 2005). A. Active Sorting with Optical Forces In active sorting techniques, most of the work has been done in introducing FACS (Galbraith et al., 1999; Givan, 2001) techniques into the microfluidic regime (Fu et al., 1999, 2002; Kruger et al., 2002; Wang et al., 2005). In this instance, the mechanisms for cell identification are very similar to that of macroscopic FACS machines and the concept of hydrodynamic focusing is maintained. The main diVerence arises in the diVerent methods for cell deflection. In microfluidic FACS, deflection of cells is done in fluid using optical forces (Fig. 5). It is worth noting here that the continued need for the sheath fluid comes largely from the active nature of the technique where a single-file flow of cells is required so that one cell at a time can be analyzed and then subsequently deflected. Microfluidic versions of FACS machines have been implemented (Applegate et al., 2006; Fu et al., 1999). At the microfluidic scale level, we need to consider the fact that we are likely to have much lower throughput, are aiming to try and separate cells from microliter samples, and indeed deal with the interesting fluid mechanics (Squires

Kishan Dholakia et al.

478 Sample input

Buffer

Buffer

Analysis region

Optical switch

Waste

Collection

Fig. 5 Schematic representation of a micro-FACS system as demonstrated by Genoptix. Similar to macroscopic FACS, the sample fluid stream is focused by a sheath flow, this flow of single-file particles is then analyzed for fluorescence and scattering, a decision made and based on this decision particles will be deflected into one of two output streams. Reprinted by permission from Macmillan Publishers Ltd.: Wang, M. M., Tu, E., Raymond, D. E., Yang, J. M., Zhang, H. C., Hagen, N., Dees, B., Mercer, E. M., Forster, A. H., Kariv, I., Marchand, P. J., and Butler, W. F. (2005). Microfluidic sorting of mammalian cells by optical force switching. Nature Biotechnol., 23, 83–87, copyright (2005).

and Quake, 2005). To adapt a FACS machine into a viable microfluidic technology, one must ensure that we retain a good throughput, purity, and recovery of cells that have not been unduly stressed by the sorting process. The use of an optical force switch to divert out selected cells of interest potentially oVers a good route to enable a micro-FACS system to be created. The work by Wang et al. (2005) showed a new microfluidic implementation of sorting using such an optical switch. The work builds upon the studies of Buican et al. (1987), who showed guiding of various types of cells in a gently focused light field and introduced the notion of deflecting out cells of interest into a reservoir of choice. Cells were hydrodynamically focused into a linear flow through a detection and subsequent deflection region. The Wang cell sorter used two lasers: one at 488 nm to excite green fluorescent protein (GFP) fluorescence in cells. The presence of the desired cell type is indicated by fluorescence recorded onto a photomultiplier.

17. Cellular and Colloidal Separation Using Optical Forces

479

A 1070-nm fiber laser, further downstream, then acts as an optical-switch deflecting cells of choice into a collection reservoir. To ensure the cells remained viable, tests of the membrane were performed with trypan blue along with an examination of two indicator genes for heat and cell shock. Other active sorting techniques include simple cell identification via video microscopy with subsequent cell rearrangement using multibeam optical tweezers and techniques more similar to microfluidic FACS where cell identification is done in a microfluidic chamber before particle deflection or placement into separate laminar flows (Buican et al., 1987; Ericsson et al., 2000; Grover et al., 2000, 2001; Oakey et al., 2002; Rodrigo et al., 2002). The discrimination of FACS methods is very high and is achieved by the sensitivity of the fluorescence detection and a suYciently strong cell deflection mechanism. The sorting speed is mainly influenced by the speed of the detection unit and response of the control electronics. B. Passive Sorting with Optical Forces It is possible to use passive cell sorting techniques in conjunction with markers, such as dielectric tags, where a functionalized dielectric sphere is attached to a specific cell population. However, the trend is towards developing techniques where the need for this pretreatment step is not required. This tag-free approach allows for separation such that the sorted analyte contains cells in their untouched state, with no need to subsequently remove the tags, or to develop the tags in the first place (often the most time consuming and costly part of the cell sorting process). When sorting is achieved without attaching tags, selectivity is obtained via the intrinsic properties of the cells. These properties are sensed most often as a size diVerence, but also as a shape and/or refractive index diVerence. As a result, such tag-free sorting is limited to sorting analyte that contains cell populations which are relatively homogenous (such as blood) rather than to sorting populations that are subject to marked size variations as part of the cells life cycle (e.g., cancer cells). One form of passive optical force-based sorting takes place where some form of kinetic lock-in of particles occurs, whereby the sorting is obtained via a size- and shape-dependent polarizability in competition with a size- and shape-dependent viscous drag (Korda et al., 2002). This can be achieved either in an array of optical traps or, in a more generalized approach, in a tailored optically induced potential landscape (Ladavac et al., 2004; MacDonald et al., 2003, 2004; Pelton et al., 2004). It is also possible to obtain passive sorting without the need for flow. This can be done, for example, in a tilted washboard-like optical potential by size-sensitive optical radiation pressure (Paterson et al., 2005; Zemanek et al., 2004b) or by adding some form of size-selective optical-moving pattern (conveyor belt) to perform separation (Cheong et al., 2006; Cizmar et al., 2006; Ricardez-Vargas et al., 2006). One of the original concepts in passive optical sorting is the idea of optical chromatography as originally proposed by Imasaka (Imasaka et al., 1995; Kaneta et al., 1997) and subsequently refined by Hart and co-workers as well as others (Hart and Terray, 2003; Hart et al., 2004, 2006; Imasaka et al., 1995;

Kishan Dholakia et al.

480

Kaneta et al., 1997; Zhao et al., 2006). In this approach, the balance between the viscous drag experienced by particles in fluid flow is balanced by the radiation pressure of a counter-propagating laser beam. Figure 9 shows an example of optical chromatography being used to separate two diVerent bacterial spores. Another technique similar in concept to optical chromatography is known as optophoresis. Optophoresis uses a balance between viscous drag and the force exerted on particles by a rapidly scanning laser beam (Zhang et al., 2004). Many more techniques are in development or have been proposed such as those based around flashing optical potentials (Libal et al., 2006; Smith et al., 2007), the particle–light–particle interaction usually referred to as optical binding (Grzegorczyk et al., 2006), optical waveguides (Grujic et al., 2005), optically induced thermophoresis (Shirasaki et al., 2006), or optically induced dielectrophoresis (Chiou et al., 2005). As to whether passive or active sorting is the most appropriate is dependent on the cells that are being investigated. In general, passive sorting is more flexible as it easily allows for sorting both with and without tags and is simple to combine with a microfluidic system, but this technique is somewhat unproven in the cellular regime. Active sorting, however, has a more established record: the body of work already done in the flow cytometry community means that active techniques like FACS bring with them a lot of valuable experience and established protocols. In the next two sections, we explore experimental implementations of sorting using optical fields: we focus on methods that are primarily passive but that may be implemented with an active element (akin to fluorescent-activated cell sorting methods), if desired.

V. Flow-Free Optical Methods In this section, we will explore the use of optical fields to separate objects in the absence of any fluid flow. Without any flow particles are usually trapped but if we make the trap potential shallow we create a metastable state such that these particles may in fact escape due to thermal activation from the optical potential well. This may in turn be exploited to separate particles and cells. In this section we look at experiments that have used static optical energy landscapes for sorting small volumes of analyte. As indicated on the earlier diagram (Fig. 2), we will thus be looking at regions 1 and 3. The Bessel light field has proven to be an interesting form of optical landscape for micromanipulation in recent years and is a good example of flow-free optical sorting. The Bessel beam is a solution of the Helmholtz equation that exhibits the property of propagation invariance: that is the intensity at a given plane is exactly the same for an idealised version of this beam as at any other plane (McGloin and Dholakia, 2005). This means it is in some sense ‘‘diVraction free.’’ This has been exploited for long range optical guiding of microparticles. The zeroth order of such a beam consists of a central bright spot surrounded by a series of concentric rings. The central spot and all of the rings each contain equal amounts of power. Interestingly, if one places colloidal

17. Cellular and Colloidal Separation Using Optical Forces

481

and cellular particles into the outer rings of the light field, one sees that the periodicity of a light pattern makes a profound diVerence with regard to the equilibrium position for a given object. This is coherent with the above-described size eVect. To employ this eVect, the landscape was tilted by slight modification of the incident beam so that this tilt biased the motion of the objects to the center of the ‘‘nondiVracting’’ Bessel mode. The objects that pass through the rings have a dependency on the height of the barrier that they have to overcome. A standard Gaussian beam optical tweezer oVers a harmonic potential well, from which thermal hopping to adjacent wells can be exploited if two or more focused beams are close to each other. As an example, in a dual tweezers system, the dynamics of particle jumps between them has been studied (Simon and Libchaber, 1992) and they proved that these jumps can be boosted if the depth of both potential wells is modulated with a period two times longer than the Kramers time (Babic et al., 2004). This can also be applied to extended optical potential landscapes such as the Bessel beam, which also have intriguing features relating to equilibria positions for trapped objects that depend on the physical parameters. Notably, objects that are large compared to the optical corrugation respond to the extended envelope of the optical field, while smaller particles respond to the individual rings within this optical landscape. The Bessel beam was used for blood sorting. Equal concentrations of both mononuclear cells and erythrocytes were suspended in an appropriate culture medium supplemented with fetal calf serum and mixed together in a sample chamber. The Bessel beam used had a 5.0-mm core size and a laser power ranging up to 800 mW. At low powers (up to 300 mW), the majority of cells were transported slowly toward the central core of the Bessel beam, where they are finally trapped by forming a vertical stack at the top of the sample chamber. At higher powers (>400 mW), the biconcave-shaped erythrocytes move toward the central core but before reaching the center, align vertically in the outer rings of the Bessel beam. Once reoriented in this manner, the erythrocytes are locked into the specific ring and are guided upward within that ring. In contrast, the spherically shaped lymphocytes move directly toward the central Bessel core, where they form a vertical stack along the central maximum as described earlier, responding rather to the overlying potential and not being locked within any ring of the Bessel beam. As the white cells are collected into the center of the beam, they experienced an upward propulsion from the optical radiation pressure from the center of the Bessel beam which thus separated the lymphoctyes and erythrocytes (Fig. 6). A judiciously placed capillary was able to extract out the lymphoctyes from the sample. The addition of colloidal beads as markers may enhance this type of sorting. Streptavidin-coated silica microspheres of 5 mm in diameter were attached to a T cell subpopulation of mononuclear cells via a mouse CD2 primary antibody and a secondary, biotinylated, antimouse antibody attachment. Attaching silica microspheres aims to enhance this passive method by selection using the beads. They are targeted to a specific subpopulation of cells via antibody–antigen binding. The cells with microspheres attached reacted to the optical landscape more strongly than cells without any attached beads due to the higher refractive index mismatch and

Kishan Dholakia et al.

482 Beam off A

Beam on (t = 0) B

t = 13 s C

t = 150 s E

Cells collected F

10 mm t = 80 s D

Fig. 6 Sorting of white and red blood cells in a Bessel beam. Red blood cells flip and align to the rings of the Bessel beam becoming trapped in the outer rings. White blood cells move along the gradient of the envelope of the Bessel function such that they are dragged into the center of the Bessel beam and guided away down its central core.

accompanying optical forces. A separation of these T cells from the ensemble of unlabeled cells was achieved in this manner. A diVerent type of static sorting without the fluid flow that is not based on the thermally activated jumps over a barrier combines light-induced particle flow in opposite directions, if they settle at appropriate parts of a set of bright fringes (Zemanek, 2004b). This type of sorting is especially useful for objects comparable or bigger than the fringe-spacing employed. Experimental separation of 2- and 5-m m, or 5- and 7-mm polystyrene beads was demonstrated. The sorting speed is proportional to the used laser power as this parameter dictates localization of particles within the fringe and increases the transport velocity along the fringe by light-induced particle flow. The stochastic thermal motion is less dominant in these cases too, and one attains higher sorting precision and discrimination (Fig. 7). A productive way to enhance optical separation without flow would be to invoke some sort of motional light pattern. In the absence of any flow, this should enhance the throughput of the sorting method while retaining much of the simplicity of the technique (region 2 in Fig. 2). Naturally, the motional speed of the pattern is the critical parameter to control because it establishes the balance of forces such that objects will follow the pattern and will jump over an inter-fringe barrier. Within this remit, a vibrating fringe pattern has been used to move and separate colloidal particles by Ricardez-Vargas et al. (2006). They used a Mach Zender interferometer type arrangement to generate a sinusoidal fringe pattern that they

17. Cellular and Colloidal Separation Using Optical Forces

483

100

Z (mm)

80 60 40 20 0 0

80

160 240 320 400 480 560 640 720 800 880 960 1040 1120 1200 Time (ms)

Fig. 7 Example of sorting of 2- and 5-mm objects in a three-beam configuration. The monosize collection of smaller spheres is assembled at the lower part of the figure following the direction of the interference fringes. Bigger spheres are pushed upward.

then projected into the sample plane of the trap. One of the mirrors in the interferometer vibrated using a piezoelectric mount and caused the fringe pattern to oscillate in a saw tooth-like manner in one direction. The size eVect was employed to separate 2- and 5-mm polystyrene particles. Moreover, they were also able to separate particles of the same size (5 mm) made of polystyrene and silica, thus showing sorting based on refractive index variations. The sorting speed is dictated by the balance between the optical force pushing the objects forward and the Stokes force resisting this movement, so the higher laser power and shorter fringe distance increases the sorting speed. The discrimination is influenced by the system’s ability to move one particle size—to create deep enough groove to suppress the thermally activated jumps between fringes—and keep the other sizes unaVected. Again with increasing laser power and shorter fringe distance, smaller diVerences in particle sizes can be separated. In recent work, Cizmar et al. (2006) have demonstrated selection and motion of submicron-sized particles near a surface in a moving periodic light pattern created by counter-propagating and interfering beams near a prism surface. This type of optical geometry has proved interesting for large-scale arrangement of microparticles and large area coverage. An added attribute is the fact that no high numerical aperture optics are used and one may freely access and image the particles from above the prism. The motion of the moving standing wave is based on the time variable phase shift between the interfering waves and may be implemented in a number of ways: by an axially movable mirror or introducing a small frequency shift between both beams (e.g. using an angular Doppler shift) with this diVerence yielding the speed of the pattern. They obtained sorting so that the pattern moved in one direction while the whole landscape was tilted in the other direction by a higher incident power coming from

Kishan Dholakia et al.

484 40

350 nm

350 nm 30 Z (mm)

350 nm 410 nm

20 10 410 nm

410 nm

0 0

0.6 1.2 1.8 2.4

3

0

0.6 1.2 1.8 2.4 Time (s)

3 0

0.6 1.2 1.8

Fig. 8 Principle (left) and examples (middle, right) of sorting of submicrometer-sized colloids of diVerent sizes diVering by just 60 nm. Polystyrene beads were used of diameters 350 and 410 nm, the smaller ones are delivered by the moving pattern like in a conveyor belt in the positive direction of the z-axis, the larger beads are insensitive to the periodic modulation of the potential and therefore they fall down due to the tilt of the periodic potential.

one of the incident beams on the prism. Thus, an object confined within the periodic structure followed the pattern motion, while a larger object responds to the envelope of the pattern and thus falls down along the potential landscape. The sorting speed and discrimination follow the same rules as above (Fig. 8). A more advanced method by which to generate 2D or 3D landscapes that are dynamically reconfigurable is to use the spatial light modulators (SLMs) (Curtis et al., 2002; Eriksen et al., 2002). They work as a phase grating (hologram) where each pixel can be independently addressed so that it has diVerent optical thickness or converting amplitude to phase using the generalized phase contrast method (Rodrigo et al., 2005). This tool can generate motional periodic light patterns with a direct application to the flow-free partile motion termed optical peristalsis (Koss and Grier, 2003). Due to the dynamic properties, the SLM also oVers sorting based on optical ratchets. To obtain a ratchet, a sort of asymmetry has to be present in the system—either asymmetric but periodic potentials or a symmetry breaking time sequence of symmetric potentials (Reimann, 2002). Lee and Grier (2005) observed particle flux reversal in a symmetry breaking time sequence of optical traps where diVerent particles experience diVerent potentials and possess diVerent diVusion coeYcients. This arrangement can also be used for thermally activated sorting. Time-dependent optical potential energy landscapes can also lead to sorting via flashing of the laser source rather than spatial scanning of the landscape (Libal et al., 2006; Smith et al., 2007). The frequency and phase of this motion is locked to the oscillation of the optical signal used to produce the landscape, leading to a ratchet eVect that gives spatial separation between strongly and weakly interacting particles. It is also possible to achieve the same ratchet-like behavior with a DC flow added orthogonal to the AC motion. This incarnation leads to two spatially separated flows of particles.

485

17. Cellular and Colloidal Separation Using Optical Forces

VI. Optical Methods with Flow One of the most promising methods to emerge in recent years has been the combination of microfluidic flow and optical forces where the flow may drive a batch of particles or cells over the potential energy landscape (regions 2 and 4 in Fig. 2). We begin by looking at the method of optical chromatography which fits into this scheme.

A. Optical Chromatography Chromatography is a separation technique that takes advantage of the diVerences in partitioning behavior between a mobile phase and a stationary phase to separate the components in a given mixture and is a commonly known technique in biology and chemistry. Optical chromatography is one of the earliest forms of optical sorting with a fluid flow and was first demonstrated over a decade ago (Imasaka et al., 1995). In this method, one uses a weakly focused laser beam in a counter-propagating fluid flow. As the particles flow along the channel, they experience a radiation pressure force from scattering of the light field and are pushed toward the focal region of the beam where they attain the highest velocities. As one might expect, this scattering force that creates the guiding diVers depending on the size and refractive index of the particle in question. In the presence of the fluid flow, the radiation pressure (guiding) force pushes the particles against the fluid flow and an equilibrium arises between the competing optical and fluid forces creating regions where particles are held. The distance beyond the focal region of the guiding laser for a given particle is known as its retention distance Z. The guiding force is related to the size of the particle and its refractive index—as stated earlier, this latter physical property is linked to its inherent chemical or biological composition. The method may be used as a powerful analytical tool and the aim is to use it for separation of a diverse range of biological materials such as blood cells, bacteria, yeast cells, pollen even bacterial warfare agents (e.g., Bacillus anthracis). If successful, this method could indeed be used for portable biological warfare detectors. Researchers in the United States are now employing this technique in a microfluidic environment (Fig. 9). As stated, the separation in this method occurs due to the balance between optical forces with Stokes forces from the fluid. The optical radiation pressure force may be given by: Frad-pressure ¼

2n1 P  a 2  Q c o

ð3Þ

where P is the laser power, n1 is the refractive index of the medium, o is the beam radius, c is the speed of light, a is the sphere radius, and Q* is a conversion factor (Hart and Terray, 2003; Hart et al., 2004) that denotes the eYciency of radiation pressure transfer from the light to the trapped object. The retention distance Z is

Kishan Dholakia et al.

486 A

B

C

Fig. 9 An example of optical chromatography demonstrated by Hart and colleagues at NRL. Here the interplay between radiation pressure and fluid drag has led to diVerent equilibrium positions for two diVerent spores. Images of (A) B. anthracis and (B) B. thuringiensis spores optically retained individually, and (C) optically retained simultaneously. The liquid flow was from right to left and the laser was propagating from left to right. Bright spots are due to laser light scatter from the spore (black rings used to highlight position). The laser focal point was positioned in the center of the main channel, 206 mm to the right of the inlet channel edge, seen in the upper left corner of each image. The scale bar represents 100 mm. Reprinted by permission from American Chemical Society: Hart, S. J., Terray, A., Leski, T. A., Arnold, J., and Stroud, R. (2006). Discovery of a significant optical chromatographic diVerence between spores of Bacillus anthracis and its close relative, Bacillus thuringiensis. Anal. Chem. 78, 3221–3225, copyright (2006).

determined by equating this formula, given in (3), to the Stokes drag in the fluid. Bacterial and fungal spores as well as various cell types have been separated using optical chromatography. Mulberry pollen and a larger ragweed pollen were held with a diVerence in retention distances of over 2 mm. The work in this field is on-going and hopes to exploit more subtle diVerences in biological samples, for example yielding diVerent retention distances for diVerent bacterial strains.

17. Cellular and Colloidal Separation Using Optical Forces

487

B. Separation Using Microfluidic Flow Over Periodic Optical Energy Landscapes In contrast to optical chromatography which relies on radiation pressure equating with Stokes drag, other forms of sorting rely on the interplay between gradient forces and viscous drag in a fluid flow. As described earlier, it is possible to achieve kinetically locked-in states when colloidal matter flows across an angled optical potential energy landscape (Korda et al., 2002). These states can lead to spatial separation of diVerent particle species as the flow velocity at which a kinetically locked-in state collapses (where the particles are no longer deflected) is lower for weakly interacting species as compared to those with a large interaction strength (Ladavac et al., 2004; MacDonald et al., 2003, 2004; Pelton et al., 2004). The key to this technique is to make the path of least resistance across the landscape diVerent for diVerent particle species. Figure 10 illustrates the sorting mechanism. When a particle approaches the optical landscape, it can do one of four things: become trapped in a local intensity maximum, be deflected by local intensity maxima, responds to the envelope function of the landscape, or go through essentially unimpeded. For sorting, one requires some particles to be deflected by hopping or channeling between local trapping sites along a diagonal of the potential landscape, while others are only weakly deflected or experience no net deflection at all. This will clearly lead to spatial separation of colloid and cells according to size, shape, and refractive index. The movement of particles through the landscape can be facilitated by introducing light channels between local trapping sites or going away from the idea of traps entirely and using angled optical guides (e.g., linear interference fringes). Two major experiments in 2003 and 2004 (Ladavac et al., 2004; MacDonald et al., 2003) have shown sorting of colloidal particles in 3D optical lattices and arrays of holographically generated optical traps. Separation angles as high as 45  have been experimentally shown and even low-index particles such as ultrasound contrast agents have been sorted. One physical incarnation of the microfluidic optical sorting approach consists of a landscape placed within a microfluidic flow chamber. The chamber brings together

Fig. 10 Sorting in an optical lattice. Two species of particles enter the lattice flowing from left to right. The weakly interacting species (dark) flows straight through while the path of least resistance for the strongly interacting species (light) is at 45  to the flow, leading to physical separation of the two species. This idea can be expanded to more than two species.

488

Kishan Dholakia et al.

two flows and then subsequently reseparates them. One flow contains a polydisperse mixture of particles, the other pure solvent or buVer solution. Strongly interacting species are deflected into the pure flow by the optical landscape and subsequently removed from the mixed flow when the two flows separate. Although originally proposed for arrays of optical traps, the technique can be extended to the more generalized idea of a potential energy landscape that can be either 2D or 3D. In fact, the sorting behavior depends strongly on the type of landscape used and how it is made. When being produced holographically, near arbitrary 2D landscapes can be produced, but such landscapes are essentially limited to two dimensions unless more complex beam generation algorithms are used. Typically if a landscape is so restricted, only particles in a single plane can be sorted. This is also the case if the landscape is produced by scanning a single laser trap such as with an acousto-optic modulator,where a time-shared array is produced. However, some of the most powerful sorting to date has been shown when using an optical landscape defined using a scanned beam (Milne et al., 2007), where as many as 4 different particle species have been separated simultaneously. To produce a truly 3D landscape, the simplest method is multibeam interference (although interference can only produce symmetric patterns) (MacDonald et al., 2003). 3D landscapes do not require that particles be well confined with a single plane of flow but can instead sort throughout a volume. This has practical benefits even when particles are all denser than their carrier medium, as it removes the crucial alignment of particle flow with optical field that may be required with 2D landscapes. By allowing for continuous particle throughput and spatially separated flows of sorted particles, this approach gives many advantages over other optical forcebased separation techniques. In the particle size range of cells, sorting with this technique is deterministic such that very high purity and eYciency can be achieved and because particles can be flowed through the landscape in parallel (rather than in single file like micro-FACS), it is easily scaleable as long as there is suYcient laser power available. When sorting cells or other colloidal objects by flowing them over an optical potential landscape, the particle density plays an important role in the behavior of the particles. Importantly, at higher densities, the streams of strongly interacting particles that are being channeled at an angle to the flow can mechanically deflect weakly interacting species. This eVect will clearly lead to an error in the output of the sorter. One approach to reducing this problem without having to go to lower input density of particles is to introduce a motional scan in the optical landscape, which corresponds to region 4 in Fig. 2. This scanning eVect leads to strongly interacting species being removed more quickly from the sorting area, increasing the nearest neighbor spacing, reducing the frequency of particle–particle interactions, and reducing the number of weakly interacting particles that are deflected (Smith et al., 2007). This regime is least studied from an experimental viewpoint, though there are some notable theoretical predictions in this area (Reichhardt and Reichhardt, 2004).

17. Cellular and Colloidal Separation Using Optical Forces

489

VII. Dielectrophoresis for Microfluidic Sorting In parallel with the controlled motion of particles using optical fields, gradients for particle and cell motion may be generated not optically but electrically. The dielectrophoretic (DEP) force is generated by the interaction of an applied electric AC field and the induced electric dipoles in neutral particles (Pohl, 1978). If the applied field is uniform, the two Coulomb forces on the charges on both sides of the particle are equal and opposite and cancel each other out. If the field is nonuniform, however, the Coulomb forces on either side of the particle will not be equal. The resulting force diVerential leads to a motion of the particle and is referred to as the DEP force. An analytic expression of this force shows that the DEP force is proportional to the gradient of the AC field E squared (Jones, 1995), Fdep ¼ 2pa3 em Re½K  ðoÞ▽ðE 2 Þ

ð4Þ

with a the particle radius, em the permittivity of the surrounding medium, and K*(o) the Clausius–Mosotti (CM) factor, which depends on the polarizability of the particle and the medium as well as the frequency of the applied AC field; Re [K*(o)] has a value between 1 and 1/2. A positive CM factor means that particles are attracted toward higher fields and vice versa. Standard dielectrophoresis has been used to sort cells based on their diVering intrinsic DEP response (Becker et al., 1995; Cheng et al., 1998; Gascoyne and Vykoukal, 2002), but this has potential problems. Just as with optical forces, cell phenotypes or diVerent target cells may show near equivalent intrinsic DEP responses making it diYcult to sort them. However, recent work has shown how this may be circumvented (Hu et al., 2005). Cells were harvested and mixed with a biotinylated monoclonal antibody. The cells were incubated with streptavidincoated polymer beads which attached themselves to these cells. Mixtures thus contained both cells with beads attached and cells without beads attached. The rare target cells in this study were Escherichia coli cells with a specific surface peptide antigen that is recognized by a monoclonal antibody. To create the appropriate electric fields for dielectrophoresis, a quadrupole electrode device was microfabricated using electron beam lithography of gold and titanium onto a glass substrate. The forces on the E. coli cells varied dramatically depending on whether they were labeled or not labeled. Tagged cells had forces of 368 pN, whereas unlabeled cells had on 57 pN exerted on them. The microfluidic chamber and these forces thus permitted one to readily deflect the labeled cells but not the unlabeled ones. A buVer flow was introduced to the center of the microfluidic chamber and though it parallels the notion of the sheath flow in FACS, in this case it was present for slightly diVerent reason: FACS systems employ the sheath flow to surround the cell flow and to reduce shear stress therein, whereas in this DEP scheme the buVer flow was in the center of the flow stream and essentially once the electrodes were powered, selected cells were readily deflected into this central region and went onto the collection region.

Kishan Dholakia et al.

490

A major issue with microfabricated FACS described earlier is the relatively low throughput: in contrast, the rather high forces combined with the relatively fast flow rates in this DEP experiment, flows of 3 mm/s near the electrodes and a flow rate of 300 ml/h. This meant that for a single round trip for a given batch of cells, a 200-fold enrichment was observed and a throughput of 10,000 cells per second was recorded, both very high compared to typically what one might expect at this small size scale. It will be interesting to see how well this active cell sorting method works for other cell types and whether such prelabeling (marking) of cells proves to be the method of choice. We note in the passive optical schemes that such marking (making them active) has yet to be fully explored—though the work using the Bessel beam has shown some interesting data. Dielectrophoresis was realized using optical control of the AC electric field (Chiou et al., 2005), as shown in Fig. 11. As a result, trapping in this geometry was demonstrated with up to a 105-fold lower power requirement than in conventional optical tweezers requiring only 10 nW/mm2. This may help realize many optical manipulation eVects hitherto inaccessible with conventional optical micromanipulation. In this section we describe this recent development and the data achieved to date. The fluid is sandwiched between two indium tin oxide (ITO)-coated glass carriers, across which an AC electric field is applied. The amorphous silicon (a-Si) layer acts as a photoconductor that exhibits high resistivity in the absence of illumination. With the laser beam on, electron–hole pairs are generated in the a-Si and the field now drops across the liquid. The a-Si layer, typically 1-mm thick, is coated with a thin film (20 nm) of silicon dioxide to prevent electrolysis. The areas of the cell where DEP forces are at play is therefore entirely controlled by the illuminating beam. The strength of the force, however, is controlled by the applied AC field, and especially its gradient [Eq. (2)]. The gradient depends on the thickness and nature of the a-Si layer (e.g., the diVusion length of the photogenerated electron–hole pairs) which is a key element in determining the smallest spot size, and therefore the resolution, that can be achieved. Calculations show that the diVusion length may readily be less than 500 nm and thus in essence it is the ability of focusing the light to a small spot size that is the key criteria here. Such a cell is made by using ITO-coated glass substrates, sputtering the silicon layer and the thin SiO2 onto the top surface and mounting them together using, for example, polystyrene spheres as spacers. Since the microfluidic functionality is derived entirely from the pattern projected onto the surface, no further lithographic patterning is necessary. Figure 11 shows some data for this type of sorting.

VIII. Conclusions There is little doubt that techniques to sort, enrich or isolate small cell populations as well as colloidal samples will be of interest to biologists, material scientists, and chemists alike. Bulk sorting methods exist and are certainly an established

491

17. Cellular and Colloidal Separation Using Optical Forces

A

D 24

Conveyor Sorting path

Dynamic wedge

z

Number of patients

Joint

20

24 m m

16 12 8 4 0

B

10 m m

0

5 10 15 20 25 30 35 40 45 50 z position (m m)

C

24 m m

10 m m

24 m m

10 mm

Fig. 11 An example of an integrated virtual optical machine. (A) Integration of virtual components, including an optical sorter path, conveyors, joints, and a wedge. The motion of diVerent components is synchronized. (B and C) Two polystyrene particles with sizes of 10 and 24 mm pass through the sorter path and are fractionated in the z direction owing to the asymmetrical optical patterns. The particle trajectories can be switched at the end of the sorter path by the optical wedge. (D) Optical sorting repeatability test. The white and black loops in B and C represent the particle traces after 43 cycles. The trace broadening at the white bar has a standard deviation of 0.5 mm for the 10-mm bead and 0.15 mm for the 24-mm bead. Reprinted by permission from Macmillan Publishers Ltd.: Chiou, P. Y., Ohta, A. T., and Wu, M. C. (2005). Massively parallel manipulation of single cells and microparticles using optical images Nature 436, 370–372, copyright (2005).

technology: these include fluorescence activated cell-sorting and techniques using magnetic beads attached to cell populations. However, there is the open question as to how we may perform sorting for rare cell types or in situations where we have very small number of cells in the first instance and indeed attain a reasonable purity and throughput for the microfluidic system in light of the challenging dimensions and physics of such laminar flow and low Reynolds number. In this respect, a portable cell sorting methodology based on optics would be highly desirable: there are no doubt prospects for such sorting and some pilot studies have shown promise particularly when it comes to lymphocytes and erythrocytes but all optical microfluidic sorting in a passive scheme remains largely unproven

Kishan Dholakia et al.

492

for a wide variety of cells. Both optical methods and DEP schemes with added surface markers (e.g., attached spheres) are certainly viable ways by which to separate cells in a microfluidic environment. The recently developed hybrid method of light-induced dielectrophoresis may possibly oVer the best of both worlds, the reconfigurability of light, yet at vastly reduced power levels, and the power of DEP forces over a large area. It is certainly a dynamic and challenging time in the field of cellular and colloidal separation. Acknowledgments We thank colleagues in all of our groups for useful discussions on the topics of separation and sorting and Dr. Frank Gunn-Moore for reading the chapter. We acknowledge support from the UK Engineering and Physical Sciences Research Council, Scottish Higher Education Funding Council, and the ATOM-3D network (contract number 508952) funded under the NEST Program of the European Commission framework 6 program, PZ acknowledges the support of the Centre of Modern Optics (LC06007) under the Ministry of Education, Youth, and Sports of the Czech Republic and the ISI Institutional research plan (AV0Z20650511).

References Applegate, R. W., Squier, J., Vestad, T., Oakey, J., Marr, D. W. M., Bado, P., Dugan, M. A., and Said, A. A. (2006). Microfluidic sorting system based on optical waveguide integration and diode laser bar trapping. Lab. Chip 6(3), 422–426. Ashkin, A., and Dziedzic, J. M. (1987). Optical trapping and manipulation of viruses and bacteria. Science 235(4795), 1517–1520. Babic, D., Schmitt, C., Poberaj, I., and Bechinger, C. (2004). Stochastic resonance in colloidal systems. Europhys. Lett. 67(2), 158–164. Becker, F. F., Wang, X. B., Huang, Y., Pethig, R., Vykoukal, J., and Gascoyne, P. R. C. (1995). Separation of human breast-cancer cells from blood by diVerential dielectric aYnity. Proc. Natl. Acad. Sci. USA 92(3), 860–864. Buican, T. N., Smyth, M. J., Crissman, H. A., Salzman, G. C., Stewart, C. C., and Martin, J. C. (1987). Automated single-cell manipulation and sorting by light trapping. Appl. Opt. 26(24), 5311–5316. Cheng, J., Sheldon, E. L., Wu, L., Heller, M. J., and O’Connell, J. P. (1998). Isolation of cultured cervical carcinoma cells mixed with peripheral blood cells on a bioelectronic chip. Anal. Chem. 70(11), 2321–2326. Cheong, F. C., Sow, C. H., Wee, A. T. S., Shao, P., Bettiol, A. A., van Kan, J. A., and Watt, F. (2006). Optical travelator: Transport and dynamic sorting of colloidal microspheres with an asymmetrical line optical tweezers. Appl. Phys. B-Lasers Opt. 83(1), 121–125. Chiou, P. Y., Ohta, A. T., and Wu, M. C. (2005). Massively parallel manipulation of single cells and microparticles using optical images. Nature 436(7049), 370–372. Cizmar, T., Siler, M., Sery, M., Zemanek, P., Garce´s-Cha´vez, V., and Dholakia, K. (2006). Optical sorting and detection of sub-micron objects in a motional standing wave. Phys. Rev. B 74, 035105. Curtis, J. E., Koss, B. A., and Grier, D. G. (2002). Dynamic holographic optical tweezers. Opt. Commun. 207(1–6), 169–175. Ericsson, M., Hanstorp, D., Hagberg, P., Enger, J., and Nystrom, T. (2000). Sorting out bacterial viability with optical tweezers. J. Bacteriol. 182(19), 5551–5555. Eriksen, R. L., Daria, V. R., and Gluckstad, J. (2002). Fully dynamic multiple-beam optical tweezers. Opt. Express 10(14), 597–602. Fu, A. Y., Chou, H. P., Spence, C., Arnold, F. H., and Quake, S. R. (2002). An integrated microfabricated cell sorter. Anal. Chem. 74(11), 2451–2457.

17. Cellular and Colloidal Separation Using Optical Forces

493

Fu, A. Y., Spence, C., Scherer, A., Arnold, F. H., and Quake, S. R. (1999). A microfabricated fluorescence-activated cell sorter. Nature Biotechnol. 17(11), 1109–1111. Galbraith, D. W., Anderson, M. T., and Herzenberg, L. A. (1999). Flow cytometric analysis and FACS sorting of cells based on GFP accumulation. In ‘‘Methods in Cell Biology,’’ Vol. 58, pp. 315–341. Academic Press Ltd., London. Gascoyne, P. R. C., and Vykoukal, J. (2002). Particle separation by dielectrophoresis. Electrophoresis 23(13), 1973–1983. Givan, A. L. (2001). ‘‘Flow Cytometry: First Principles.’’ Wiley-Liss, USA. Gleeson, J. P., Sancho, J. M., Lacasta, A. M., and Lindenberg, K. (2006). Analytical approach to sorting in periodic and random potentials. Phys. Rev. E 73(4), Art. No. 041102 Part 1. Grier, D. G. (2003). A revolution in optical manipulation. Nature 424(6950), 810–816. Grover, S. C., Gauthier, R. C., and Skirtach, A. G. (2000). Analysis of the behaviour of erythrocytes in an optical trapping system. Opt. Express 7(13), 533–539. Grover, S. C., Skirtach, A. G., Gauthier, R. C., and Grover, C. P. (2001). Automated single-cell sorting system based on optical trapping. J. Biomed. Opt. 6(1), 14–22. Grujic, K., Helleso, O. G., Hole, J. P., and Wilkinson, J. S. (2005). Sorting of polystyrene microspheres using a Y-branched optical waveguide. Opt. Express 13(1), 1–7. Grzegorczyk, T. M., Kemp, B. A., and Kong, J. A. (2006). Stable optical trapping based on optical binding forces. Phys. Rev. Lett. 96(11), Art. No. 113903. Hart, S. J., and Terray, A. V. (2003). Refractive-index-driven separation of colloidal polymer particles using optical chromatography. Appl. Phys. Lett. 83(25), 5316–5318. Hart, S. J., Terray, A., Kuhn, K. L., Arnold, J., and Leski, T. A. (2004). Optical chromatography of biological particles. Am. Lab. 36, 13–17. Hart, S. J., Terray, A., Leski, T. A., Arnold, J., and Stroud, R. (2006). Discovery of a significant optical chromatographic diVerence between spores of Bacillus anthracis and its close relative, Bacillus thuringiensis. Anal. Chem. 78(9), 3221–3225. Herzenberg, L. A., Parks, D., Sahaf, B., Perez, O., Roederer, M., and Herzenberg, L. A. (2002). The history and future of the fluorescence activated cell sorter and flow cytometry: A view from Stanford. Clin. Chem. 48(10), 1819–1827. Hu, X. Y., Bessette, P. H., Qian, J. R., Meinhart, C. D., Daugherty, P. S., and Soh, H. T. (2005). Marker-specific sorting of rare cells using dielectrophoresis. Proc. Natl. Acad. Sci. USA 102(44), 15757–15761. Imasaka, T., Kawabata, Y., Kaneta, T., and Ishidzu, Y. (1995). Optical chromatography. Anal. Chem. 67(11), 1763–1765. Jones, T. B. (1995). Electromechanics of Particles. Cambridge University Press ISBN-13: 978-0521431965. Kaneta, T., Ishidzu, Y., Mishima, N., and Imasaka, T. (1997). Theory of optical chromatography. Anal. Chem. 69(14), 2701–2710. Korda, P. T., Taylor, M. B., and Grier, D. G. (2002). Kinetically locked-in colloidal transport in an array of optical tweezers. Phys. Rev. Lett. 89(12), 128–301. Koss, B. A., and Grier, D. G. (2003). Optical peristalsis. Appl. Phys. Lett. 82(22), 3985–3987. Kramers, H. A. (1940). Brownian motion in the field of force and diVusion model of chemical reactions. Physica 7(4), 284–304. Kruger, J., Singh, K., O’Neill, A., Jackson, C., Morrison, A., and O’Brien, P. (2002). Development of a microfluidic device for fluorescence activated cell sorting. J. Micromech. Microeng. 12(4), 486–494. Lacasta, A. M., Sancho, J. M., Romero, A. H., and Lindenberg, K. (2005). Sorting on periodic surfaces. Phys. Rev. Lett. 94(16), 188–902. Ladavac, K., Kasza, K., and Grier, D. G. (2004). Sorting mesoscopic objects with periodic potential landscapes: Optical fractionation. Phys. Rev. E 70(1), Art. No. 010901 Part 1. Lee, S. H., and Grier, D. G. (2005). Flux reversal in a two-state symmetric optical thermal ratchet. Phys. Rev. E 71(6), Art. No. 060102 Part 1. Lekner, J. (2005). Force on a scatterer in counter-propagating coherent beams. J. Opt. A: Pure Appl. Opt. 7, 238–248.

494

Kishan Dholakia et al. Libal, A., Reichhardt, C., Janko, B., and Reichhardt, C. J. O. (2006). Dynamics, rectification, and fractionation for colloids on flashing substrates. Phys. Rev. Lett. 96(18), Art. No. 188301. MacDonald, M. P., Neale, S., Paterson, L., Riches, A., Dholakia, K., and Spalding, G. C. (2004). Cell cytometry with a light touch: Sorting microscopic matter with an optical lattice. J. Biol. Regulat. Homeostat. Agents 18(2), 200–205. MacDonald, M. P., Spalding, G. C., and Dholakia, K. (2003). Microfluidic sorting in an optical lattice. Nature 426(6965), 421–424. McGloin, D., and Dholakia, K. (2005). Bessel beams: DiVraction in a new light. Contemp. Phys. 46(1), 15–28. Milne, G., Rhodes, D., MacDonald, M., and Dholakia, K. (2007). ‘‘Fractionation of polydisperse colloid with acousto-optically generated potential energy landscapes.’’ Opt. Lett. 32, 1144–1146. Molloy, J. E., Dholakia, K., and Padgett, M. J. (2003). Preface: Optical tweezers in a new light. J. Mod. Opt. 50(10), 1501–1507. Neuman, K. C., and Block, S. M. (2004). Optical trapping. Rev. Sci. Instrum. 75(9), 2787–2809. Oakey, J., Allely, J., and Marr, D. W. M. (2002). Laminar-flow-based separations at the microscale. Biotechnol. Progress 18(6), 1439–1442. Paterson, L., Papagiakoumou, E., Milne, G., Garce´s-Cha´vez, V., Tatarkova, S. A., Sibbett, W., GunnMoore, F. J., Bryant, P. E., Riches, A. C., and Dholakia, K. (2005). Light-induced cell separation in a tailored optical landscape. Appl. Phys. Lett. 87, Art. No. 123901. Pelton, M., Ladavac, K., and Grier, D. G. (2004). Transport and fractionation in periodic potentialenergy landscapes. Phys. Rev. E 70(3), Art. No. 031108 Part 1. Pohl, H. A. (1978). ‘‘Dielectrophoresis: The Behaviour of Neutral Matter in Nonuniform Electric Fields.’’ Cambridge University Press, Cambridge. Prasad, P. N. (2003). ‘‘Introduction to Biophotonics.’’ John Wiley and Sons Inc., Hoboken, New Jersey. Reichhardt, C., and Reichhardt, C. J. O. (2004). Directional locking eVects and dynamics for particles driven through a colloidal lattice. Phys. Rev. E 69(4), Art. No. 041405 Part 1. Reimann, P. (2002). Brownian motors: Noisy transport far from equilibrium. Phys. Rep. 361, 57–265. Ricardez-Vargas, I., Rodriguez-Montero, P., Ramos-Garcia, R., and Volke-Sepulveda, K. (2006). Modulated optical sieve for sorting of polydisperse microparticles. Appl. Phys. Lett. 88(12), Art. No. 121116. Rodrigo, P. J., Daria, V. R., and Gluckstad, J. (2005). Dynamically reconfigurable optical lattices. Opt. Express 13(5), 1384–1394. Rodrigo, P. J., Eriksen, R. L., Daria, V. R., and Gluckstad, J. (2002). Interactive light-driven and parallel manipulation of inhomogeneous particles. Opt. Express 10(26), 1550–1556. Shirasaki, Y., Tanaka, J., Makazu, H., Tashiro, K., Shoji, S., Tsukita, S., and Funatsu, T. (2006). On-chip cell sorting system using laser-induced heating of a thermoreversible gelation polymer to control flow. Anal. Chem. 78(3), 695–701. Siler, M., Cizmar, T., Sery, M., and Zemanek, P. (2006). Optical forces generated by evanescent standing waves and their usage for sub-micron particle delivery. Appl. Phys. B-Lasers Opt. 84(1–2), 157–165. Simon, A., and Libchaber, A. (1992). Escape and Synchronization of a Brownian Particle. Phys. Rev. Lett. 68(23), 3375–3378. Smith, R. L., Spalding, G. C., Dholakia, K., and MacDonald, M. P. (2007). Colloidal sorting in dynamic optical lattices. J. Opt. A: Pure. Appl. Opt., special issue on optical manipulation. (in press). Squires, T. M., and Quake, S. R. (2005). Microfluidics: Fluid physics at the nanoliter scale. Rev. Mod. Phys. 77(3), 977–1026. Tatarkova, S. A., Sibbett, W., and Dholakia, K. (2003). Brownian particle in an optical potential of the washboard type. Phys. Rev. Lett. 91(3), Art. No. 038101. Wang, M. M., Tu, E., Raymond, D. E., Yang, J. M., Zhang, H. C., Hagen, N., Dees, B., Mercer, E. M., Forster, A. H., Kariv, I., Marchand, P. J., and Butler, W. F. (2005). Microfluidic sorting of mammalian cells by optical force switching. Nature Biotechnol. 23(1), 83–87.

17. Cellular and Colloidal Separation Using Optical Forces

495

Zelenina, A. S., Quidant, R., Badenes, G., and Nieto-Vesperinas, M. (2006). Tunable optical sorting and manipulation of nanoparticles via plasmon excitation. Opt. Lett. 31(13), 2054–2056. Zema´nek, P., Jonas, A., Sramek, L., and Liska, M. (1998). Optical trapping of Rayleigh particles using a Gaussian standing wave. Opt. Commun. 151, 273–285. Zema´nek, P., Jonas, A., Jakl, P., Jezek, J., Sery, M., and Lisˇka, M. (2003). Theoretical comparison of optical traps created by standing wave and single beam. Opt. Commun. 220, 401–412. Zema´nek, P., Jonas, A., and Liska, M. (2002). Simplified description of optical forces acting on a nanoparticle in the Gaussian standing wave. J. Opt. Soc. Am. A. 19, 1025–1034. Zema´nek, P., Karasek, V., and Sasso, A. (2004a). Optical forces acting on Rayleigh particle placed into interference field. Opt. Commun. 240(4–6), 401–415. Zema´nek, P., Karasek, V., and Sery, M. (2004b). Behaviour of colloidal microparticles in a planar 3-beam interference field. Proc. SPIE 5514, 15–26. Zhang, H. C., Tu, E., Hagen, N. D., Schnabel, C. A., Paliotti, M. J., Hoo, W. S., Nguyen, P. M., Kohrumel, J. R., Butler, W. F., Chachisvillis, M., and Marchand, P. J. (2004). Time-of-flight optophoresis analysis of live whole cells in microfluidic channels. Biomed. Microdevices 6(1), 11–21. Zhao, B. S., Koo, Y. M., and Chung, D. S. (2006). Separations based on the mechanical forces of light. Anal. Chim. Acta 556(1), 97–103.